I am making offline analysis of the hands I own after playing sometime online.
I play Omaha only, but this may apply to Texas Holdem too.
I want to generate with code notes of my opponents and insert them on pokertracker database with code.
To simplify it, let's say I want to generate notes for flush projects on flop as:
- Bets a little flush project Flop.
- Bets quite flush project Flop.
- Bets a lot flush project Flop.
I want to know, if my oponnet is raising, how much probably is beting with flush project. That can help me make decissions as playing top double pairs on a flop with two suited cards. If my oponnent never raises with flush project and is raising I can put him on a set and fold my two pair. And viceversa I can pay with my double top pair if a third suited card don't comes.
I know the best here is to use machine learning algorithms, but I don't have so much programing and math knowledges. I want just to aproximate with an index, plot an histogram of the frequency of index of all my oponnents and decide where I put the limits .
The database I generated from hands stores as a boolean (true or false) value:
- Call to bet.
- Call to raise.
- Bets.
- Raises.
I cannot calculate VPIP or AFQ since my data is for hands with showdown. When my oponnents folded I don't know wich cards they had.
To generate my index this are the values I give presently to each play:
- Call to bet. (3 points)
- Call to raise. (5 points)
- Bets. (5 points)
- Raises. (10 points)
Do you think the points I give to each play are accurated, or I should change them?